Ambiguity discrimination for ship detection using Sentinel-1 repeat acquisition operations

This paper presents a framework to identify recurrent targets at sea in Sentinel-1 images making use of the repeat acquisition operations of this sensor and time series analysis of the images collected. Recurrent targets are those targets that regularly appear in the same location over different acquisitions. They can broadly be classified in real fixed structures (e.g. oil platforms), which are likely to appear in all the images, and ambiguities of fixed targets, which appear in a given location only for a specific observation geometry and for a specific set of sensor parameters. Ambiguities traditionally pose a serious challenge for ship detection systems, as they often result in false alarms. This framework can be used to discriminate real targets (fixed or moving) from ambiguities, and its main strength lies in its practicality, inasmuch as it can be used with single and dual-polarisation, complex and non-complex Synthetic Aperture Radar (SAR) images. It can potentially be used with any orbital SAR sensor, but it is more easily exploited for Sentinel-1 thanks to the free and open data policy and the pre-established observation plan of this sensor.